Programming Social Processes with Action Languages
A social process is a joint of activities (business, poitical, administrative and so on) performed in a social context. It can be a computational society when a :SocialMiddleware is provided and it can be programmed as a first-class social connector (:SocialInteraction). The abstract machine for the SOcialMiddleware they use action languages (C+): a declarative language to specify structure and dynamics, based upon a propositional, non-monotonic, causal logic, and Ccalc tool (SAT solver).
The social middleware is defined through a set of sorts, objects, fluents, actions, variables and axioms. But to reduce the complexity of these specifications they make a sort-oriented specification. He explains using a tennis match example how the society can be modelled.
WIth all that, a programming language for agent societies can be defined (SPEECH: a societal programming language – www.speechlang.org).
Representing and Reasoning about Norm-Governed Organisations with Semantic Web Languages
SW languages are more limited, but they offer a stardard language (open), advantadges in reasoning (decidability and optimized reasoners). They want to capture roles and role classification, institutionalised powers, normative systems, violations and temporal relationships (hey…. just as I want!!! I have to read it carefully). OWL and SWRL extended for temporal relations (SWRLTab editor) for before, after and during (–> converted into PDDL).
They define permissions, abbligations and permision for roles and inheritance of norms. You can override them if you need more especific norms. There’re general rules form norms (ex. if an act is permitted then it’s not prohibited / if an act is obligatory the it’s permitted… ) Other cocept modelled is «power» (a role has the powerto execute an action). Norms can have conditions and deadlines, but the DL reasoner has to be extended to del with temporal relationships (this can be an interesting work to do) To detect violations, the system can check if an agent triesto perform an act that is prohibited. That can be detected automatically by the reasoner.
But we can reason an inconsistent ontology by tolerating contradictions (that’s interesting for open systems or self-organising systems)
Limitations (comparing with event calc.): only binary operators, static knowledge, no modifications
On Partial Deductions and Conversational Agents
Many times, human conversations include more information that a simple yes/no answer, so if agents want to interact with human complex conversation schemes have to be included in their mental state.
The mental state of an agent is formed by a ….., a …. and a set of rules. Goals are facts that agents want to be solve. THe input and the output interface are set of facts. Agents communicate using an own simple speech act messages. Interface engine os based on rle specialisation uses an approximate reasoning context and the true value is not binary, but a fuzzy one.
Examples to show the mentak state cycle and hoe agents communicates to complete their information.